R Notebook

#install.packages("knitr")
#install.packages("grid")
#install.packages("MLRMPA")
#install.packages("dprep")
#install.packages("normalr")
#install.packages("ggcorrplot")

library(gridExtra)
library(dplyr)
library(lubridate)
library(magrittr)
library(ggplot2)
library(tidyr)
library(knitr)
library(normalr)
library(ggcorrplot)
#library(MLRMPA)
#??src_mysql
my_db <- src_mysql(
  dbname = "covid",
  host = "localhost",
  user = "root",
  password = "1234"
)
my_db

src:  mysql 8.0.21 [root@localhost:/covid]
tbls: citytemperature, covid_confirmed_yearly, covid_deaths_yearly, covid_recovered_yearly, covid19_confirmed_global, covid19_deaths_global,
  covid19_recovered_global, full_datas, gdp, gdp19, gdp2019, healthranking, imf-country, owid-covid-data, population, temp2019, world_temp

##import data
df_conf <- tbl(my_db, sql("select * from covid_confirmed_yearly"))
df_conf <- as.data.frame(df_conf)
df_conf

df_deaths <- tbl(my_db, sql("select * from covid_deaths_yearly"))
df_deaths <- as.data.frame(df_deaths)
df_deaths

df_recover <- tbl(my_db, sql("select * from covid_recovered_yearly"))
df_recover <- as.data.frame(df_recover)
df_recover

##check the time frame of the data
n.col <- ncol(df_conf)
dates <- names(df_conf)[5:n.col]%>% mdy()
range(dates)

[1] "2020-01-22" "2021-01-11"

min.date <- min(dates)
max.date <- max(dates)
min.date.txt <- min.date %>% format('%d %b %Y')
max.date.txt <- max.date %>% format('%d %b %Y')

#clean data
cleanData <- function(data) {
  ## remove some columns
  data %<>% select(-c(Province.State, Lat, Long)) %>% rename(country=Country.Region)
  ## convert from wide to long format
  data %<>% gather(key=date, value=count, -country)
  ## convert from character to date
  data %<>% mutate(date = date %>% mdy())
  ## aggregate by country
  data %<>% group_by(country, date) %>% summarise(count=sum(count, na.rm=T)) %>% as.data.frame()
  return(data)
}
## clean the three data sets
data.confirmed <- df_conf %>% cleanData() %>% rename(confirmed=count)

`summarise()` regrouping output by 'country' (override with `.groups` argument)

data.deaths <- df_deaths %>% cleanData() %>% rename(deaths=count)

`summarise()` regrouping output by 'country' (override with `.groups` argument)

data.recovered <- df_recover %>% cleanData() %>% rename(recovered=count)

`summarise()` regrouping output by 'country' (override with `.groups` argument)

data <- data.confirmed %>% merge(data.deaths, all=T) %>% merge(data.recovered, all=T)
data

## countries/regions with confirmed cases, excl. cruise ships
countries <- data %>% pull(country) %>% setdiff('Cruise Ship')
data

data.world <- data %>% group_by(date) %>%
  summarise(country='World',
            confirmed = sum(confirmed, na.rm=T),
            deaths = sum(deaths, na.rm=T),
            recovered = sum(recovered, na.rm=T))

`summarise()` ungrouping output (override with `.groups` argument)

data %<>% rbind(data.world)
data

data %<>% mutate(current.confirmed = confirmed - deaths - recovered)
View(data)

#rate
data %<>% arrange(country, date)
n <- nrow(data)
day1 <- min(data$date)
data %<>% mutate(new.confirmed = ifelse(date == day1, NA, confirmed - lag(confirmed, n=1)),
                 new.deaths = ifelse(date == day1, NA, deaths - lag(deaths, n=1)),
                 new.recovered = ifelse(date == day1, NA, recovered - lag(recovered, n=1)))
data %<>% mutate(new.confirmed = ifelse(new.confirmed < 0, 0, new.confirmed),
                 new.deaths = ifelse(new.deaths < 0, 0, new.deaths),
                 new.recovered = ifelse(new.recovered < 0, 0, new.recovered))
## death rate based on total deaths and recovered cases
data %<>% mutate(rate.upper = (100 * deaths / (deaths + recovered)) %>% round(1))
## lower bound: death rate based on total confirmed cases
data %<>% mutate(rate.lower = (100 * deaths / confirmed) %>% round(1))
## death rate based on the number of death/recovered on every single day
data %<>% mutate(rate.daily = (100 * new.deaths / (new.deaths + new.recovered)) %>% round(1))
View(data)

## convert from wide to long format
data.long <- data %>%
  select(c(country, date, confirmed, current.confirmed, recovered, deaths)) %>%
  gather(key=type, value=count, -c(country, date))
## set factor levels to show them in a desirable order
data.long %<>% mutate(type=recode_factor(type, confirmed='Total Confirmed',
                                         current.confirmed='Current Confirmed',
                                         recovered='Recovered',
                                         deaths='Deaths'))
View(data.long)

##Number of case World
world <- filter(data.long,country == 'World')
plot1 <- world %>% filter(type != 'Total Confirmed') %>%
  ggplot(aes(x=date, y=count)) +
  geom_area(aes(fill=type), alpha=0.5) +
  labs(title=paste0('Numbers of Cases Worldwide - ', max.date.txt)) +
  scale_fill_manual(values=c('red', 'green', 'black')) +
  theme(legend.title=element_blank(), legend.position='bottom',
        plot.title = element_text(size=7),
        axis.title.x=element_blank(),
        axis.title.y=element_blank(),
        legend.key.size=unit(0.2, 'cm'),
        legend.text=element_text(size=6),
        axis.text=element_text(size=7),
        axis.text.x=element_text(angle=45, hjust=1))
plot2 <- world %>%
  ggplot(aes(x=date, y=count)) +
  geom_line(aes(color=type)) +
  labs(title=paste0('Numbers of Cases Worldwide (log scale) - ', max.date.txt)) +
  scale_color_manual(values=c('purple', 'red', 'green', 'black')) +
  theme(legend.title=element_blank(), legend.position='bottom',
        plot.title = element_text(size=7),
        axis.title.x=element_blank(),
        axis.title.y=element_blank(),
        legend.key.size=unit(0.2, 'cm'),
        legend.text=element_text(size=6),
        axis.text=element_text(size=7),
        axis.text.x=element_text(angle=45, hjust=1)) +
  scale_y_continuous(trans='log10')
## show two plots side by side
grid.arrange(plot1, plot2, ncol=2)

## Current Confirmed Cases
data.world <- data %>% filter(country=='World')
n <- nrow(data.world)
plot1 <- ggplot(data.world, aes(x=date, y=current.confirmed)) +
  geom_point() + geom_smooth() +
  xlab('') + ylab('Count') + labs(title='Current Confirmed Cases') +
  theme(axis.text.x=element_text(angle=45, hjust=1))
plot2 <- ggplot(data.world, aes(x=date, y=new.confirmed)) +
  geom_point() + geom_smooth() +
  xlab('') + ylab('Count') + labs(title='Daily New Confirmed Cases') +
  theme(axis.text.x=element_text(angle=45, hjust=1))
## show two plots side by side
grid.arrange(plot1, plot2, ncol=2)

## a scatter plot with a smoothed line and vertical x-axis labels
plot1 <- ggplot(data.world, aes(x=date, y=deaths)) +
  geom_point() + geom_smooth() +
  xlab('') + ylab('Count') + labs(title='Accumulative Deaths') +
  theme(axis.text.x=element_text(angle=45, hjust=1))
plot2 <- ggplot(data.world, aes(x=date, y=recovered)) +
  geom_point() + geom_smooth() +
  xlab('') + ylab('Count') + labs(title='Accumulative Recovered Cases') +
  theme(axis.text.x=element_text(angle=45, hjust=1))
plot3 <- ggplot(data.world, aes(x=date, y=new.deaths)) +
  geom_point() + geom_smooth() +
  xlab('') + ylab('Count') + labs(title='New Deaths') +
  theme(axis.text.x=element_text(angle=45, hjust=1))
plot4 <- ggplot(data.world, aes(x=date, y=new.recovered)) +
  geom_point() + geom_smooth() +
  xlab('') + ylab('Count') + labs(title='New Recovered Cases') +
  theme(axis.text.x=element_text(angle=45, hjust=1))
## show four plots together, with 2 plots in each row
grid.arrange(plot1, plot2, plot3, plot4, nrow=2)

## convert from wide to long format, for drawing area plots
rates.long <- data %>%
  select(c(country, date, rate.upper, rate.lower, rate.daily)) %>%
  gather(key=type, value=count, -c(country, date))
# set factor levels to show them in a desirable order
rates.long %<>% mutate(type=recode_factor(type, rate.daily='Daily',
                             rate.upper='Upper bound'))

## ranking by confirmed cases
data.latest.all <- data %>% filter(date == max(date)) %>%
  select(country, date,confirmed, new.confirmed, current.confirmed,
         recovered, deaths, new.deaths, death.rate=rate.lower) %>%
  mutate(ranking = dense_rank(desc(confirmed)))
#View(data.latest.all)
k <- 20
## top 20 countries: 21 incl. 'World'
top.countries <- data.latest.all %>% filter(ranking <= k + 1) %>%
  arrange(ranking) %>% pull(country) %>% as.character()
top.countries %>% setdiff('World') %>% print()

 [1] "US"             "India"          "Brazil"         "Russia"         "United Kingdom" "France"         "Turkey"         "Italy"         
 [9] "Spain"          "Germany"        "Colombia"       "Argentina"      "Mexico"         "Poland"         "Iran"           "South Africa"  
[17] "Ukraine"        "Peru"           "Netherlands"    "Indonesia"     

data.latest <- data.latest.all %>% filter(!is.na(country)) %>%
  mutate(country=ifelse(ranking <= k + 1, as.character(country), 'Others')) %>%
  mutate(country=country %>% factor(levels=c(top.countries, 'Others')))
data.latest %<>% group_by(country) %>%
  summarise(confirmed=sum(confirmed), new.confirmed=sum(new.confirmed),
            current.confirmed=sum(current.confirmed),
            recovered=sum(recovered), deaths=sum(deaths), new.deaths=sum(new.deaths)) %>%
  mutate(death.rate=(100 * deaths/confirmed) %>% round(1)) 

`summarise()` ungrouping output (override with `.groups` argument)

data.latest

data.latest %<>% select(c(country, confirmed, deaths, death.rate,
                          new.confirmed, new.deaths, current.confirmed,recovered)) %>%
  mutate(recover.rate=(100 * recovered/confirmed) %>% round(1))
data.latest

df_pop <- tbl(my_db, sql("select * from population "))
df_pop <- as.data.frame(df_pop)
df_pop <- rename(df_pop,"country"="Country")

# Add World Population
world_pop <- sum(df_pop$`Population (2020)`)
df_pop[nrow(df_pop) + 1,] = c("World", world_pop)

# Add Other Countries Population
top_pop <- filter(df_pop, df_pop$country %in% top.countries & df_pop$country != "World")
top_pop <- sum(top_pop$`Population (2020)` %>% as.numeric())
others_pop <- (world_pop - top_pop) 
df_pop[nrow(df_pop) + 1,] = c("Others", others_pop)
#View(df_pop)

data.latest <- merge(x = data.latest, y = df_pop, by = "country", all.x = TRUE) 
data.latest

data.latest <- rename(data.latest,"population" = "Population (2020)")
data.latest$population <- data.latest$population %>% as.numeric()
data.latest  <- data.latest %>%
  select(c(country, confirmed, deaths, death.rate,
                          new.confirmed, new.deaths,
                          current.confirmed, recovered, recover.rate, population)) %>%
  mutate(confirm.rate = (100 * confirmed / population) %>% round(1))
data.latest

data.latest %>% mutate(death.rate=death.rate %>% format(nsmall=1) %>% paste0('%'))

NA
NA

## convert from wide to long format, for drawing area plots
data.latest.long <- data.latest %>% filter(country!='World') %>%
  gather(key=type, value=count, -country)
## set factor levels to show them with proper text and in a desirable order
data.latest.long %<>% mutate(type=recode_factor(type,
                                                confirmed='Total Confirmed',
                                                deaths='Total Deaths',
                                                death.rate='Death Rate (%)',
                                                new.confirmed='New Confirmed (compared with one day before)',
                                                new.deaths='New Deaths (compared with one day before)',
                                                current.confirmed='Current Confirmed',
                                                recover.rate = 'Recover Rate(%)',
                                                confirm.rate = 'Confirmed Rate(%)'))
#View(data.latest.long)
data.one.dem <- filter(data.latest.long,type=='Total Confirmed'
                       | type=='Total Deaths'
                       | type=='Current Confirmed')
data.two.dem <- filter(data.latest.long,type=='Death Rate (%)'
                       | type=='New Confirmed (compared with one day before)'
                       | type=='New Deaths (compared with one day before)'
                       | type=='Recover Rate(%)'
                       | type=='Confirmed Rate(%)')
data.two.dem

## bar chart
data.one.dem %>% ggplot(aes(x=country, y=count, fill=country, group=country)) +
  geom_bar(stat='identity') +
  geom_text(aes(label=count, y=count), size=3, vjust=0) +
  xlab('') + ylab('') +
  labs(title=paste0('Top 20 Countries with Most Confirmed Cases - ', max.date.txt)) +
  scale_fill_discrete(name='Country', labels=aes(count)) +
  theme(legend.title=element_blank(),
        legend.position='none',
        plot.title=element_text(size=13),
        axis.text=element_text(size=8),
        axis.text.x=element_text(angle=45, hjust=1)) +
  facet_wrap(~type, ncol=1, scales='free_y')

data.two.dem %>% ggplot(aes(x=country, y=count, fill=country, group=country)) +
  geom_bar(stat='identity') +
  geom_text(aes(label=count, y=count), size=3, vjust=0) +
  xlab('') + ylab('') +
  labs(title=paste0('Top 20 Countries with Most Confirmed Cases - ', max.date.txt)) +
  scale_fill_discrete(name='Country', labels=aes(count)) +
  theme(legend.title=element_blank(),
        legend.position='none',
        plot.title=element_text(size=13),
        axis.text=element_text(size=8),
        axis.text.x=element_text(angle=45, hjust=1)) +
  facet_wrap(~type, ncol=1, scales='free_y')

##GDP
df_gdp <- tbl(my_db, sql("select * from gdp"))
df_gdp <- as.data.frame(df_gdp)
df_gdp <- rename(df_gdp,"country"="Real GDP growth (Annual percent change)")
df_gdp <- select(df_gdp,c("country","2012","2013","2014","2015","2016","2017","2018","2019","2020","2021"))
df_gdp

df_gdp2019 <- tbl(my_db, sql("select * from gdp19"))
df_gdp2019 <- as.data.frame(df_gdp2019)
df_gdp2019

NA

#healthranking
df_healt <- tbl(my_db, sql("select * from healthranking"))
df_healt <- as.data.frame(df_healt)
df_healt <- select(df_healt,c("country","healthCareIndex"))
View(df_healt)

#temp
df_temp <- tbl(my_db, sql("select * from temp"))

Error in .local(conn, statement, ...) : 
  could not run statement: Table 'covid.temp' doesn't exist

#Top 20 with gdp
data.longGDP <- df_gdp %>% gather(key=year, value=GDP, -c(country))
data.top <- data.latest %>% filter(country!='World'& country!='Others')
#data.top <- head(data.top,20)
View(data.top)
data.gdp <- filter(data.longGDP,year=='2020')
#View(data.gdp)
#merge
mergcountry = function(data1,data2){
  data <- merge(x = data1, y = data2, by = "country", all.x = TRUE) 
  return(data)
}
data.top.world <- merge(x = data.top, y = df_gdp2019, by = "country", all.x = TRUE) %>% 
  select(-c(code,rank,new.confirmed,new.deaths,current.confirmed,population)) %>% 
  rename(GDP="GDP (millions of US dollars)")
data.top.world <- merge(x = data.top.world, y = df_healt, by = "country", all.x = TRUE) %>%
  rename(healthcare="healthCareIndex")
#data.top.world <- mergcountry(data.top.world, df_temp)
View(data.top.world)
normalize = function(data){
  #return ((data - min(data,na.rm = TRUE))/(max(data,na.rm = TRUE) - min(data,na.rm = TRUE)))
  z <- scale(data);
  tanh(z/2)
}
norm_data = as.data.frame(apply(data.top.world[,2:9],2,normalize))
corr_data <- norm_data
norm_data$country <- c("Argentina","Bangladesh","Brazil","Chile","Colombia","France","Germany","India","Iran","Italy","Mexico","Pakistan","Peru","Russia","saudi Arabia","South Africa","Spain","Turkey","United Kingdom","US")
View(norm_data)
norm_data_plot <- select(norm_data,"country","confirm.rate","death.rate","recover.rate","healthcare","GDP")
norm_data_plot %<>% gather(key=type, value=count, -c(country))
level_order <- factor(norm_data_plot$type, 
                      level = c("GDP","healthcare","recover.rate","death.rate","confirm.rate"))
ggplot(data = norm_data_plot, aes(x=country, y=level_order, fill=count)) + 
  geom_tile() +
  scale_fill_gradient(low = "pink", high = "blue") +
  xlab("") +
  ylab("") +
  theme_bw() +
  theme(axis.text.x = element_text(angle = 90,vjust = 1))+
  theme(
    axis.line = element_blank(),
    axis.ticks = element_blank(),
    panel.grid.minor = element_blank(),
    panel.grid.major = element_blank(),
    panel.border = element_blank(),
    panel.background = element_blank(),
    #legend.position = "none"
  )

NA

#rank GDP
data.top.hight <- data.gdp %>% select(country, year,GDP) %>%
  mutate(ranking = dense_rank(desc(GDP)))
data.top.hight

k <- 15
top.gdp <- data.top.hight %>% 
  #filter(ranking <= k + 1) %>% 
  arrange(ranking)
top.gdp <- head(top.gdp,21)
data.top.low <- data.gdp %>% select(country, year,GDP) %>%
  mutate(ranking = dense_rank(GDP))
low.gdp.long <- data.top.low %>% 
  #filter(ranking <= k + 1) %>% 
  arrange(ranking)
View(low.gdp.long)
low.gdp <- head(low.gdp.long,23)
low.gdp

#correlation
corr_data %<>% select(c(GDP,confirm.rate,death.rate,recover.rate,healthcare))
head(corr_data)

cor(corr_data)

                    GDP confirm.rate death.rate recover.rate healthcare
GDP           1.0000000    0.4297936 -0.2360041   -0.3765615  0.2518394
confirm.rate  0.4297936    1.0000000 -0.3345825   -0.6868148  0.4543381
death.rate   -0.2360041   -0.3345825  1.0000000    0.1626707 -0.1470020
recover.rate -0.3765615   -0.6868148  0.1626707    1.0000000 -0.7549792
healthcare    0.2518394    0.4543381 -0.1470020   -0.7549792  1.0000000

ggcorrplot(cor(corr_data),hc.order = TRUE,
           outline.color = "white",
           colors = c("#6D9EC1","white","#E46726"),
           lab = TRUE)